How to Make an AI Avatar Video with HeyGen (Script to Export) 2026

You’ve got a script and no time to film it.

To make an AI avatar video with HeyGen, pick an avatar and paste your script into short scenes. Choose a voice, preview the audio free, render once, then download the MP4. Free proves the idea. Creator at $29 is where watermark-free 1080p starts.

We checked every step and rate against HeyGen’s help center in October 2026.

The catch is the engine. On the same plan, one minute of HeyGen avatar video costs 4 credits or 48.

What do you need before you open AI Studio?

Disclosure: the HeyGen and ElevenLabs links here are affiliate links. If you upgrade through them, we may earn a commission. The steps and warnings stay the same.

• A HeyGen account. Free covers a test: up to three one-minute videos a month (fewer in some regions), watermarked.
• A finished script. Plan on about 150 words per minute. A 60-second explainer is roughly 150 words, not 400.
• A voice plan. A stock HeyGen voice, a clone, or your ElevenLabs voice through an API key.

Still comparing tools? Our AI avatar generator roundup puts HeyGen next to its rivals.

Which avatar engine should you pick?

Pick the engine before the face, because the engine decides your bill. HeyGen runs three avatar models, and each charges per minute of generated video.

Engine	Photo look	Video look	Best for
Avatar III	7 credits/min	4 credits/min	Training videos, bulk
Avatar IV	16 credits/min	31 credits/min	Photo avatars, stylised characters, custom motion
Avatar V	48 credits/min	48 credits/min	Your digital twin, realistic presenters

Avatar V is HeyGen’s default for real-human avatars and digital twins, and it needs a video-based look. Avatar IV is the default for photo avatars and non-human characters. Avatar III is the budget engine.

Is Avatar V worth 12 times the Avatar III rate?

For the sales video with your own face on it, probably. For a 20-minute onboarding course, no. Nobody sitting through a compliance module will thank you for sharper eyelids.

Our take: start on Avatar IV with a photo look at 16 credits a minute. It sits mid-range on price, and HeyGen built it for photo avatars. Move to Avatar V once you have a digital twin worth showing off.

How do you make your first HeyGen avatar video?

The whole job runs in AI Studio in six moves. Credits only get spent at move five.

Move five is where first videos go wrong. People render the full script, hear a mangled brand name in scene three, fix it, and pay to render that scene again.

Write a script the avatar can actually say

A script that reads well can still sound wrong out loud. Text-to-speech trips on the same things every time, so fix them before you render.

• Numbers and times. Write them as you’d say them: 01:18 becomes one minute and eighteen seconds, 3/8 becomes three eighths.
• Brand names and acronyms. Double-click the word, choose Pronunciation and type it phonetically (a-eye for AI). That fixes one instance only. Brand Glossary fixes the whole project.
• Pauses. Put the cursor where a breath belongs and click Pause. Length adjusts in half-second steps.
• One language per script. A Hindi phrase inside an English script can come out garbled. Give each language its own scene or project.

Drafting in a doc? Upload it from the top-left menu. Each line of a .txt file becomes a scene, and anything over 1,000 characters gets split. Save Word files as .docx; .doc won’t load.

Keep scenes short. A typo fix then re-renders 15 seconds, not the whole video.

Which voice should your avatar use?

Use a stock voice for the first video, then upgrade once the format works. The Voice panel offers four routes: the library, a new clone, an AI-designed voice, or a third-party voice.

Under the hood, voices run on engines such as ElevenLabs, Panda (with Voice Director for tone and pacing) and Starfish for Chinese, Japanese and Korean. Leave it on Auto unless a voice sounds off.

Bringing your ElevenLabs voice into HeyGen

Already cloned your voice in ElevenLabs? Don’t clone it twice. Our ElevenLabs voice cloning guide covers the clone itself. Connecting it takes one API key:

• In AI Studio open Voice, then + New Voice, Integrate 3rd Party Voice, Import Voice.
• Choose ElevenLabs and paste your ElevenLabs API key.
• Seeing Invalid API key? Edit the key in ElevenLabs and allow Text to Speech, Voices, Voice Generation, Models, History and User.

Worth paying for both tools? Only when the voice is the point of the video. Our ElevenLabs review shows where its voices beat built-in ones, and ElevenLabs has a free tier to test your clone. For a weekly product update, HeyGen’s own clone on Creator is enough.

Want the avatar to be you?

Record at least two minutes of continuous footage of yourself talking, then a short consent clip. That’s a digital twin, and even Free includes one slot.

• Film at 1080p and 30fps or better. A phone works. Tap and hold your face on screen to lock focus and exposure.
• Speak in the footage, even if you’ll use another voice. HeyGen learns lip sync from your mouth moving to real speech.
• Close your lips between sentences. Open with a two-to-three-second pause, keep head turns under 30 degrees and hands below your chest.
• Don’t cut the take. HeyGen wants one continuous shot. Fluff a line? Keep talking or restart.

The consent clip must run under 30 seconds, read word for word by the same person. Making a twin of a colleague? They can record it on their own phone by scanning a QR code.

What does one minute of avatar video really cost?

Credits cost about 5 cents each on every self-serve plan, so your engine sets the price, not your plan.

Creator is $29 for 600 credits, or $0.048 each. Pro starts at $49 for 1,000 ($0.049), and Business top-ups are $5 per 100.

One-minute video	Credits	Cost on Creator	Per month on Creator
Avatar III, video look	4	about $0.19	150 videos
Avatar IV, photo look	16	about $0.77	37 videos
Avatar IV, video look	31	about $1.50	19 videos
Avatar V	48	about $2.32	12 videos

Those monthly counts assume zero re-renders. Redo one 20-second Avatar V scene twice and you’ve spent 32 extra credits, two-thirds of another finished minute.

Custom Expressive Motion on Avatar IV adds 40 credits a minute. A 30-second scene with custom gestures costs 20 credits more than the plain version.

Creator can’t buy extra credits. Run dry and you upgrade to Pro, paying the difference. Unused monthly credits roll over one cycle, and annual plans keep them until renewal. Our HeyGen pricing breakdown compares every tier.

Real-world tip: Free runs on a video quota, not credits. Use it to test faces and voices. Upgrade once you know which engine you’ll render on, because only then can you predict your month.

Three mistakes that waste HeyGen credits

Defaulting to Avatar V for everything. A 10-minute training video at 48 credits a minute eats 480 of Creator’s 600.

Custom motion on every scene. Save it for the hook. A talking head reading a list doesn’t need choreographed hands.

Cancelling with credits in the bank. Drop to Free and every unused credit disappears. Spend them first.

When is HeyGen the wrong tool?

When you need editing more than a presenter. Descript edits real recordings through the transcript, which beats generating a fake you for podcast clips.

When you want one expressive character clip, not a scripted presenter. Our Hedra review covers a credit-based tool built for that job.

For scripted talking heads at volume, we’d start with HeyGen, because scene-by-scene editing keeps fixes cheap.

Bottom line

Getting a HeyGen AI avatar video out the door is simple: short script, small scenes, free audio preview, one test render, then submit.

Choose the avatar engine before the face. Avatar IV photo looks suit most social videos, Avatar V suits your digital twin, and Avatar III handles long internal training. Never render a scene you haven’t heard first.

Frequently asked questions